Gabor wavelet enabled Zernike moments for effective content based image retrieval

Abstract

This paper presents a novel technique Gabor Wavelet enabled Zernike Moments (GWZMs) to improve retrieval accuracy. Zernike Moments (ZMs) extract global details of an image. In order to enable ZMs to extract local details from image, Gabor filters are applied to generate various filtered images. Afterwards, ZMs are computed for these filtered images. The performance of GWZMs is evaluated on various databases such as MPEG-7 CE-2 and COIL-100 using different similarity measures. Experimental analysis shows that GWZMs improve precision obtained by ZMs from 76.61% to 85.4% at 100% recall on COIL-100 database. The performance of GWZMs (measured as average Bulls Eye Performance (BEP)) is 63.14% as compared to 56.27% (with ZMs) for MPEG-7 CE-2. The GWZMs are tested on other applied areas also, wherein they attain average BEP of 68.48% and 61.45% in comparison to 61.52% and 58.05% with ZMs on Springer medical image and electronics books graphical databases respectively. © 2013 Springer-Verlag.